Professional Documents
Culture Documents
ATM Introduction: The Grand Unification
ATM Introduction: The Grand Unification
2005/03/11
Agenda
What is it? Who wants it? Who did it?
Header and Switching
ATM Layer Hypercube
Adaptation Layers
Signaling
Addresses
2005/03/11
What is ATM ?
High-Speed Virtual Circuits
PVC and SVC
No error recovery
2005/03/11
Design Ideas
Asynchronous TDM
ATM
copy
copy
Synchronous TDM
Fast Switching and short delays
through constant timeslots
fake
Solved through
constant frame sizes
Protocol Transparent
fake
Solved through
adaptation layers
2005/03/11
2005/03/11
Cell Switching
Forwarding of cells implemented in HW
Very fast
2005/03/11
ATM Usage
Public and private networks
LAN, MAN, WAN
2005/03/11
ATM Network
ATM DTE
ATM DTE
UNI
ATM DCE
ATM DCE
ATM DCE
NNI
ATM DCE
ATM DTE
2005/03/11
ATM DTE
8
Virtual Circuits
VPI
1/253
1/200
1/452
3/253
3/200
3/452
IN
9/99
VCI
3/2
3/2
66/6
100/6
OUT
66/6
9/99
5/88
1/321
2005/03/11
1/123
2005/03/11
10
The CCITT (ITU-T) standardization process is very time consuming because the
final result should meet all demands of all participants such as governments,
vendors, users, and other industry representatives. Because of this, the ATM
Forum was founded to accelerate the development. Although this efforts also
helped the ITU-T standardization efforts, there are important differences between
both standards.
The ATM Forum was founded in 1991.
10
2005/03/11
11
Note that both public and private networks cover SVCs but in order to establish
SVCs we need routing tables. The routing tables can be created automatically in
private ATM networks using the ATM routing protocol PNNI (Private-NNI). In
public networks the routing tables are managed manually.
PNNI is a link-state routing protocol that enables quality of service routing.
11
NNI Types
Public NNI
Private ATM
Public ATM
B-ICI
(NNI-ICI)
Public ATM
ICI...Inter Carrier Interface
(C) Herbert Haas
2005/03/11
12
12
What is B-ISDN?
ITU-T identified several demands
Emerging need for broadband services
High speed switching
Improved data- and image processing
capabilites available to the user
Support for real-time services
Support for interactive services
Support for distribution services
Circuit and packet mode
2005/03/11
13
13
2005/03/11
14
IEEE 802.6 is a MAN standard and also known as "Distributed Queued Dual
Bus". Interestingly, DQDB is very similar to ATM in many aspects (same frame
sizes, same idea of adaption layers, etc.), However this alternative, that has been
pushed by many proponents of the data communication industry, vanished from
market.
Currently (Gigabit-) Ethernet seems to replace ATM in many areas since it is
easier to deploy, easier to manage and less expensive. However, many customers
suspects its reliability and quality of service.
14
48 Byte Payload
(C) Herbert Haas
2005/03/11
5 Byte
Header
15
USA proposed a cell size of 64 bytes while the Europeans suggested 32 bytes. At
the end they met in the middle at 53 bytes.
Why are ATM cells so small at all? One important reason is to minimize
packetization delay (splitting data in cells) so that realtime applications can be
easily supported. Another reason also associated with realtime (voice)
applications is to support the implementation of echo-cancelation devices. Long
and variable packet sizes make a echo cancelator run into trouble...
15
Trailer
Datagram
Header
Trailer
Header
ATM Layer
2 Byte
Trailer
(C) Herbert Haas
44 Byte
Datagram-Segment
2005/03/11
2 Byte 5 Byte
Header Header
16
16
ATM Header
UNI Header
GFC
NNI Header
VPI
VPI
VPI
VCI
VCI
PT
HEC
8 bit VPI for users
CLP
PT
CLP
HEC
12 bit VPI inside the network
2005/03/11
17
The Generic Flow Control (GFC) field is only used on the UNI but not
transported into the network. The GFC is not used today as there are better
methods available (special flow-control cells).
The Virtual Path Identifier (VPI) is four bits longer inside the network (on NNIs)
in order to support better traffic aggregation (Virtual Path Switching).
The Payload Type (PT) is used to identify the cell payload (OAM, Resource
Management, ...)
The Cell Loss Priority (CLP) has the same meaning as the DE-bit in Frame
Relay. Using the CLP we can distinguish between important and not-soimportant cells (CLP=1). Of course we hope that the network would be so kind to
drop CLP=1 cells first in case of congestion.
The Header Error Check (HEC) is a CRC-8 to protect the header only not the
payload! You may ask how framing is accomplished? For this purpose a receiver
device has to compute the CRC-8 for each 4 bytes and look for a match with the
following byte. In case of 6 successive hits the ATM layers are synchronized.
Note: Although ATM is an asynchronous TDM technology it is actually
implemented synchronously. There are no gaps between cells but idle cells
(VPI/VCI set to zero and payload is 01010101010101010101...).
17
Payload Type
User
signaling
bit
100
OAM F5 segment
101
OAM F5 end-to-end
110
Resource Management (RM)
Also used by AAL5 to indicate end of
block (EOB)
Other combinations: user data
2005/03/11
18
18
Header Fields
Cell Loss Priority (CLP)
Similar to DE bit in Frame Relay
Identifies less important cells
2005/03/11
19
19
VC Switching
VC Switching distinguishes each
virtual circuit according to its
VPI and VCI
Many table entries necessary
73/10
10/12
27/99
3/20
19/19
20/44
1/8
5/77
80/31
2/1
53/76
22/33
4/5
17/91
10/12
21/41
112/89
40/30
(C) Herbert Haas
2005/03/11
20
20
VP and VC Switching
VC switching only when necessary
(at borders and branches)
Fewer table entries necessary
Probably faster
73/10
10/12
100/99
100/20
VPI=100
100/19
20/44
VPI=152
2/1
VPI=65
200/91
200/12
200/41
112/89
40/30
(C) Herbert Haas
2005/03/11
21
If one customer asks for another 100 virtual circuits, the provider does not need
to add 100 additional entries in every switching tableas he must do with Frame
Relay for example- because one VPI represents a number of VCIs and inside the
network VPI switching is performed.
21
Connection Types
Point-to-point:
unidirectional or bidirectional
Point-to-multipoint:
unidirectional only
2005/03/11
22
22
User Plane
Higher Layer
ATM Adaptation Layer (AAL)
ATM Layer
Additional headers
and fragmentation
according service
Create ATM
cells and
headers
Physical Layer
(C) Herbert Haas
2005/03/11
23
23
...And In Detail
(Resources, Parameters,
OAM Flow, Meta-Signaling)
Outband signaling
in designated VCs
(I-LMI)
Management Plane
Control
Signaling
and
Control
Service
Dependent
User Plane
Class A Class B Class C Class D
CBR for
Circuit
Emulation
VBR for
Audio
and Video
Connection
oriented
Data
Connection
less
Data
ATM Layer
Transmission Convergence (TC)
2005/03/11
PDH and
SONET/SDH
24
24
Control Plane
DTE
DCE
DCE
0/18 (PNNI)
0/5 (Q.2931)
Control Plane
2005/03/11
25
25
Reserved Labels
VPI
VCI
Function
0- 15
ITU-T
16 - 31
ATM Forum
0
0
0
3
Idle Cell
Segment OAM Cell (F4)
Signaling
0
0
16
17
ILMI
LANE
18
PNNI
2005/03/11
26
26
Physical Layer
Transmission Convergence (TC)
allows simple change of physical
media
PDH, SDH, SONET
HEC and cell delineation
2005/03/11
27
27
Interface Examples
Standard
SDH STM-1
PDH E4
PDH DS3
PDH E3
PDH E2
PDH J2
PDH E1
PDH DS1
SDH STM-4
SDH STM-1
SDH STM-1
SDH STM-4
SDH STM-4
SDH STM-4
SDH STM-1
SDH STM1
FIber Channel
TAXI
SONET STS1
ATM 25
Speed
155,52
139,264
44,736
34,368
8,448
6,312
2,048
1,544
622,08
155,52
155,52
622,08
622,08
622,08
155,52
155,52
155,52
100
51,84
25,6
2005/03/11
Medium
Coax
Coax
Coax
Coax
Coax
TP/Coax
TP/Coax
TP
SM fiber
SM fiber
MM fiber
SM fiber
MM (LED)
MM (Laser)
UTP5
STP (Type1)
MM fiber
MM Fiber
UTP3
UTP3
Comments
75 Ohm
75 Ohm
75 Ohm
75 Ohm
75 Ohm
110/75 Ohm
120/75 Ohm
100 Ohm
62,5 um
100 Ohm
150 Ohm
62,5 um
62,5 um
Encoding
CMI
CMI
B3ZS
HDB3
HDB3
B6ZS/B8ZS
HDB3
AMI/B8ZS
SDH
SDH
SDH
NRZ
NRZ
NRZ
NRZI
NRZI
8B/10B
4B/5B
NRZI
NRZI
Connector
BNC
BNC
BNC
BNC
BNC
RJ45/BNC
9pinD/BNC
RJ45/RJ48
SC
ST
SC
SC (ST)
SC (ST)
SC (ST)
RJ45
9pinD
MIC
RJ45
RJ45
Usage
WAN
WAN
WAN
WAN
WAN
WAN
WAN
WAN
LAN/WAN
LAN/WAN
LAN/WAN
LAN
LAN
LAN
LAN
LAN
LAN
LAN
LAN
LAN
28
28
ATM Layer
Multiplexing and demultiplexing of
cells according VPI/VCI
Switching of cells
"Label swapping"
Note: origin of MPLS
2005/03/11
29
29
Adaptation Layers
ATM only provides bearer service
ATM cannot be used directly
Applications must use adaption
layers to access the ATM layer
Consist of SAR and CS
Part of DTEs only
Transparent for switches (DCEs)
2005/03/11
30
The ATM adaption layers translate between the specific worlds of higher layer
protocols such as IP, X.25, or PCM-Voice and the cell-nature of the ATM layer
itself. Using specific adaptation layers, nearly every application can be
transported over ATM. This capability emphasizes again the B-ISDN idea that
has been realized.
Note: "Application" means simply "any higher layer communication protocol".
Just consider ATM as a "Transport Layer" (not to confuse with the OSI layer 4!)
that provides "Bearer Services".
30
Adaptation Sub-Layers
Convergence Sublayer (CS)
Service dependent functions
(clock recovery, message identification)
Adds special information
(e. g. Frame Relay header)
2005/03/11
SSCS
Service Specific CS
Application 2
SSCS
Service Specific CS
CPCS
Common Part Convergence Sublayer
31
The Convergence Sublayer (CS) is divided in two further sublayers, the Common
Part Convergence Sublayer (CPCS) and the Service Specific Convergence
Sublayer (SSCS). The CPCS is common to all instances of a specific AAL.
Therefore only one CPCS has been defined per AAL while many SSCS can be
defined for the same AAL.
31
AAL1
Constant Bit Rate (CBR)
Circuit Emulation
Expensive
Overprovisioning like leased line
necessary
Queuing prefers AAL1 cells over all
other traffic (in case of congestion)
2005/03/11
32
32
AAL1
1 byte
Header
47 byte
SAR PDU
CSI
(0)
SN
CRC
CSI
(1)
SN
CRC
CSI .
SN ..
CRC ...
P .
47 byte
SAR PDU
Pointer
46 byte
SAR PDU
2005/03/11
33
The CSI values in cells 1, 3, 5, and 7 are interpreted as a 4 bit timing value in
order to measure timing differences between network's reference clock and
transmitter clock. The 3 bit CRC only protects the 4 bit CSI+SN field. The parity
bit is set such that the parity of the 8 header bits is even.
The sequence number is basically for cell-loss detection.
33
AAL2
Analog applications that require
timing informations but not CBR
Variable Bit Rate (VBR)
Compressed audio and video
2005/03/11
34
34
2005/03/11
35
AAL0 simply means using ATM without any adaptation layer. The AAL2 CPCS
allows variable packet length between 1 byte to 45 bytes, so the packetization
delay can be kept very small, if needed. On the other hand, bandwidth efficiency
is achieved by multiplexing several AAL2 connections within one ATM
(VPI/VCI) pipe.
Note: Instead of having partially filled ATM cells, the CPCS will fill each cell
with AAL2 packets from multiple sessions!
35
AAL2 CS
1...45/64 bytes
of user data
Connection Identifier
(identifies different
AAL2 connection)
8 bit
6 bit
CID
LI
Length Indicator
(Indicates actual
payload length)
5 bit
1...45/64 bytes
of user data
5 bit
UUI HEC
User-to-user Identifier
(Some SSCS convey
a sequence number)
5 Bit CRC
to protect the header
Total length:
9...72 Bytes
ATM
Header
6 bit
1 bit 1 bit
OSF
SN P
2005/03/11
48 Bytes
36
The slide above shows the most important tasks of the Common Part
Convergence Sublayer (CPCS). Basically, CPCS-packets consist of a one-byte
header and up to 45 bytes user data (typically compressed voice or video). Even
longer payloads are supported if an appropriate SSCS is used additionally
(I.366.1).
Does make a 1-bit sequence number any sense? Yes, together with the Parity (P)
it provides a simple mechanism to check for cell loss and errors inside the "Start
Field" (SF = OSF + SN + P).
One of the most important applications for AAL2 today is compressed voice as
for GSM, which generates 260 bits packet every 20 ms. To each of these packet a
AAL2 header is attached, indicating the length and the connection. For GSM
there is no need for pre-processing by an SSCS, in fact there is no SSCS
implemented.
Although the multiplexing feature of AAL2 dramatically improves bandwidth
efficiency, the main disadvantage is a significant increase in overhead. But this
overhead is not worth to mention when comparing AAL2 with the older method
of partial filling. When users want to send more than a few hundreds of bytes per
packet then AAL5 is more efficient.
36
AAL3 + AAL4
AAL3 designed to carry
connection-oriented packets
Such as X.25 or Frame Relay
2005/03/11
37
37
AAL3/4 Step 1: CS
must match
CPI Btag BAsize
Data Packet
PAD AL Etag
4 Bytes
Len
4 Bytes
2005/03/11
38
CPI indicates the interpretation of the remaining fields. Currently only value zero
is defined which indicates that the values in BAsize and Length should be
interpreted as amount of bytes.
38
4 Bytes
44 Bytes
PAD
2 Bytes
2
10
MID
Seq
Multiplexing
Nr
Identifier
00 = COM
01 = EOM
10 = BOM
11 = SSM
ATM Cell
(C) Herbert Haas
2 Bytes
6
10
Length
CRC
Length of segment
Default: 44
Exception: EOM and SSM
63 in EOM means "Discard it!"
ATM
Header
2005/03/11
39
39
AAL3/4
Can multiplex different streams of data
on the same ATM connection
Up to 210 streams using the same VPI/VCI
2005/03/11
40
40
AAL5
Favorite for data communication
AAL 5 simulates connectionless data
interface
Allows simple migration to ATM
Smallest overhead
Convergence Layer:
8 byte trailer in last cell
SAR Layer:
just marks EOM in ATM header (PT)
(C) Herbert Haas
2005/03/11
41
AAL5 is the most widely used AAL today. Also UNI signaling, ILMI and PNNI
signaling is done upon AAL5.
41
AAL5 Segmentation
Example: IP Packet
(0-65535 Bytes)
8 Bytes
User-to-User
Identification
Only used with
MPEG
48 Bytes
LEN
CPI
UU
PAD
2 Bytes
CRC-32
Mostly
Unused
(00000000)
PAD
VPI
VCI
1 CLP
(C) Herbert Haas
2005/03/11
LEN
ATM
Header
CPI
UU
ATM
Header
CRC-32
HEC
42
The Common Part Indicator (CPI) is not really used until today, so all 8 bits are
set to zero.
Most interestingly, the length field is 16 bits that is 64 KByte packets are
supported. So, ATM is one of the few technologies that support maximum IP
packet sizes (IP also supports up to 65535 Bytes per packet).
42
IP Packet
IP Packet
IP Packet
2005/03/11
43
43
2005/03/11
44
44
Signaling
ATM Forum UNI signaling specification
UNI 3.0, 3.1 and 4.0 standardized
UNI 2.0
UNI 3.0
UNI 4.0
PVC
PVC+SVC, CBR+VBR+UBR
+ABR, QoS Negotiation
2005/03/11
45
45
Signaling Layers
Q.2931
Service Specific
Convergence Sublayer
Service Specific
Coordination
Function
SSCF
SSCS
SAAL
(Q.2130)
SSCOP
(Q.2110)
CPCS
SAR
ATM
Layer
(C) Herbert Haas
Service Specific
Connectionoriented Protocol
Common Part
Convergence
Sublayer
Signaling
AAL
2005/03/11
46
46
ATM Addresses
ATM Forum defined three addressformats
ISO DCC NSAP format
ISO ICD NSAP format
E.164 Address format
2005/03/11
47
47
ATM Addresses
Different types of ATM addresses
All have 20 byte length
All consist of three main parts
Prefix (Basically topology information)
End System Identifier (ESI)
NSAP Selector (Selects application)
20 Byte
Prefix
ESI
Sel
13 Bytes
6 Bytes
1 Byte
2005/03/11
48
The NSAP Selector field is basically the same as the port number in TCP.
48
Address Flavours
DCC ATM Address Format (AFI=39)
AA
Authority and
Format Identifier
reserved RD AREA
Administrative
Authority
ESI
Sel
ESI
Sel
ESI
Sel
Area Identifier
AA
reserved RD AREA
Routing Domain
AFI
E.164
ISDN Number
2005/03/11
RD AREA
Endsystem Identifier
NSAP Selector
49
49
Summary
ATM is the solution for B-ISDN
Different broadband services upon common
cell relay technology
Quality of Service
Details in other module
VP and VC switching
2005/03/11
50
50
Quiz
Which framing is used with XDSL?
What are the 4 ATM basic service
types regarding QoS?
ATM flow control is similar to...?
Which concepts of ATM have been
copied for IP networks?
2005/03/11
51
51
Hints
Q1: ATM cells
Q2: CBR, VBR, ABR, UBR
Q3: Frame Relay ECN
Q4: Label Swapping (MPLS), QoSSignaling (RSVP) and QoS-Marking
(DSCP)
2005/03/11
52
52